Apple just previewed a new slate of accessibility features, headlined by Personal Voice, a tool that creates a personal speech twin using 15 minutes of training audio. Designed for people at risk of losing their speaking ability, the model processes entirely on-device to keep voice data private. The features arrive later this year.
Google just teamed up with Taito to announce SPACE INVADERS: World Defense, an AR mobile game launching later this summer. Powered by ARCore and Geospatial Creator, it turns your local neighborhood, landmarks, and weather into a 3D alien shooting gallery.
Google just revealed that Android will get generative AI wallpapers this fall, letting users generate custom backgrounds via suggested prompts. The feature uses diffusion models to create original art and adapts your phone's color theme accordingly, with 3D cinematic wallpapers coming to Pixel next month.
Google just unveiled its next-gen PaLM 2 language model at I/O, powering the newly updated Bard and serving as the foundation for the company's latest AI updates. Instead of chasing pure parameter scale, PaLM 2 sharpens logic, math, and support for over 100 languages. It also spawns specialized variants like Codey and Med-PaLM 2, plus a mobile version hitting 20 tokens per second.
Google just announced the integration of WebGPU into Chrome at Google I/O, enabling web apps to tap directly into laptop and phone GPUs. The company also demoed a web-based Stable Diffusion tool generating images directly on-device. Why it matters: Shifting AI from massive cloud servers to local browsers slashes latency while keeping private data on your own machine.
Humane co-founder Imran Chaudhri just gave the world its first full look at the company’s secretive screenless wearable in a TED talk. The clip-on device operates independently of a phone, projecting calls onto your palm, summarizing emails, and translating speech using an AI voice clone. Why it matters: Humane wants to replace smartphone screens entirely, though how practical this puck is in everyday life remains an open question.
A leaked internal memo from a Google researcher claims that neither Google nor OpenAI has a moat against open source. The document argues that open-source contributors are out-innovating tech giants using efficient methods like LoRA and cheap compute. Why it matters: The era of relying solely on massive closed models may be ending as fast, nimble, community-driven alternatives close the gap.
Humane co-founder Imran Chaudhri just gave a sneak peek of the startup’s standalone AI wearable at TED. The screenless gadget operates without a smartphone, uses voice and gestures, and projects caller info directly onto your palm in green light. Chaudhri also demoed real-time French translation, with more details promised in the coming months.
Stability AI just launched the Alpha version of Stable LM, an open-source language model suite debuting at 3B and 7B parameter sizes. Trained on 1.5 trillion tokens, the lightweight models handle text and code, with base checkpoints licensed for commercial use under CC BY-SA-4.0. Why it matters: it gives developers a way to build local, private AI applications without being handcuffed to proprietary cloud APIs.
Apple just began testing natural language generation features for Siri under the codename "Bobcat" in the tvOS 16.4 beta. The feature is currently limited to telling jokes and handling timers on Apple TV, though code exists across iOS and macOS. It is aimed at upgrading Siri's decade-old template system rather than building a ChatGPT clone.
Local LLMs just became real on everyday hardware. Thanks to Georgi Gerganov’s new llama.cpp and 4-bit quantization, Meta’s LLaMA can now run directly on personal laptops like a MacBook. It brings GPT-3-class capabilities straight to individual developers, breaking AI out of the corporate cloud garden.
Ford just launched a new wholly-owned subsidiary called Latitude AI, bringing on about 550 former Argo AI staff to develop hands-free and eyes-off automated driving. The team will focus on upgrading systems like BlueCruise to Level 3 autonomy. Why it matters: Ford is officially moving away from costly robotaxi bets to focus on driver-assist tech that can actually make money today.
TikTok just went wild over a new Bold Glamour filter that looks unnervingly realistic, seamlessly tracking faces even when hands block the camera. Experts suspect mobile generative AI like GANs is running behind the scenes, though TikTok refuses to confirm. Beyond the engineering feat, the indistinguishable edits are reigniting major concerns over body dysmorphia.
Qualcomm just demoed Stable Diffusion 1.5 running entirely on a Snapdragon 8 Gen 2 smartphone, generating a 512x512 image in under 15 seconds. That shatters previous mobile records and easily beats similar tests on iOS. Running models directly on edge chips promises zero server costs and rock-solid privacy, though this custom software build isn't publicly available yet.
Nvidia just rolled out an AI "Eye Contact" feature in Broadcast 1.4 that replaces your eyes with simulated ones pointing straight at the camera. Designed for creators reading off scripts, the tool matches eye color and blinks in real time, though testers noticed occasional unnatural darting. It brings teleprompter-like convenience to anyone with an RTX graphics card.
Scientists from the University of Stirling and Hack the Planet developed a smart camera trap that runs on-device AI and connects straight to satellites. Tested in Gabon, it identifies elephants with 82% accuracy and alerts rangers in seven minutes without Wi-Fi or cellular service, helping stop poachers and prevent human-wildlife conflict.
BMW just brought out Arnold Schwarzenegger at CES to unveil the i Vision Dee, an electric sports sedan designed around next-level human-machine interaction. While "Dee" stands for digital emotional experience, the standout feature is its electronic ink exterior, which can shift between patterns across 32 different colors.
Iris Technology just launched a limited beta of webAI, a no-code platform designed to train and deploy computer vision models fast. The company claims it needs only one-fifth of the training data and one-third of the time compared to YoloV7, and runs directly on consumer laptops without cloud compute.
AMD just unveiled its Ryzen 7040 series mobile processors, marking the first time an x86 chip features a dedicated on-die AI engine called Ryzen AI. Built on the XDNA adaptive architecture, AMD claims it outperforms Apple's M2 in both AI speed and battery efficiency. The first laptops powered by the chip will ship this March.
Nvidia just unveiled RTX Video Super Resolution, bringing AI video upscaling to web browsers starting in February. Users with RTX 30 and 40 series GPUs will be able to upscale 1080p web videos to 4K inside Chrome and Edge. Why it matters: It sharpens blurry online streams into crisp 4K footage, though RTX 20 series cards and Firefox are left out for now.
